Papers by Milan Bhan
Self-AMPLIFY: Improving Small Language Models with Self Post Hoc Explanations (2024.emnlp-main)
Copied to clipboard
| Challenge: | Autoregressive Large Language Models (LLMs) have demonstrated "emergent abilities" such as in-context learning, instruction following and reasoning. |
| Approach: | They propose a method that generates rationales from post hoc explanation methods applied to small language models to improve their own performance. |
| Outcome: | The proposed method improves on four SLMs and five datasets with strong reasoning abilities. |
Towards Achieving Concept Completeness for Textual Concept Bottleneck Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a novel TCBM generator is proposed to build concept labels in unsupervised manner using a small language model. |
| Approach: | They propose a complete textual concept bottleneck model that builds concept labels in unsupervised manner using a small language model. |
| Outcome: | The proposed model achieves striking results against existing models in terms of concept basis completeness and concept detection accuracy. |